At present our CSV import dialog seems to use TEXT as the type for all columns.

It would be useful to do some form of detection on the data (after the last row is loaded?), then change the column types to match whatever is detected:

• INTEGER and REAL should both be fairly easy to detect and implement
• CSV has no defined way of handling BLOB/binary values in the official RFC, so probably no need to detect that

We may need to define something custom for binary/BLOB importing anyway, just so it can be done in the (rare?) cases it's needed. eg for importing DBHub.io CSV exports that have binary/BLOB data in them